DiagramZu MCP server

@diagramzu/mcp

MCP server for diagramzu.ai. Lets Claude Code, Claude Desktop, Cursor, Windsurf, ChatGPT custom GPTs, and any MCP client read and write Mermaid diagrams in your Space.

You author diagrams by talking to your AI — it stores them in your Space at diagramzu.ai/app/d/, where your team can read them and publish a public link.

Available in the official MCP Registry as ai.diagramzu/mcp.

1. Get a token

Sign up at diagramzu.ai, then create an API token at diagramzu.ai/app/settings/connections. Tokens look like dzlive… and are scoped to one Space — no separate space-id needed.

2. Connect your client

The hosted server is the easy path: no install, no build. Just paste a config.

Claude Code

claude mcp add --scope user --transport http diagramzu https://mcp.diagramzu.ai/mcp \
  --header "Authorization: Bearer dz_live_xxx"

Claude Desktop, Cursor, Windsurf, Cline

Add to your client's MCP config (~/Library/Application Support/Claude/claudedesktopconfig.json on macOS for Claude Desktop, ~/.cursor/mcp.json for Cursor, etc.):

{
  "mcpServers": {
    "diagramzu": {
      "type": "http",
      "url": "https://mcp.diagramzu.ai/mcp",
      "headers": { "Authorization": "Bearer dz_live_xxx" }
    }
  }
}

ChatGPT custom GPT (Actions)

In the GPT builder, add an MCP server action pointing to https://mcp.diagramzu.ai/mcp with a Bearer-token authentication header set to your dzlive… token.

Local stdio (for clients that don't speak remote MCP)

npx -y @diagramzu/mcp

with environment:

DIAGRAMZU_BASE_URL=https://diagramzu.ai
DIAGRAMZU_API_TOKEN=dz_live_xxx
DIAGRAMZU_SPACE_ID=<your space id>

Most users should prefer the remote HTTP transport above — the stdio path exists for clients without HTTP MCP support.

Tools

Which URLs are public

Two different things get called "a link" here, and only one of them works for someone outside your Space:

Shortened. The full README is on GitHub.